Guard Ring

Andreas said:

Hello

I am designing an photodiode amplifier with an OPA129. I have a question regarding the guard rings mentioned in the document http://focus.ti.com/lit/ds/symlink/opa129.pdf Figure 3 shows how to route the guard ring on the pcb, Figure 2 shows where to connect it. In my case with an inverting amplifier, I connect the guard ring to pins 8 and 3. The part I don't understand is how I get into the guard ring. Do I have to get in with a via to another layer? Or do I extend the guard ring to include the pad where I connect the photodiode to? In that case the guard ring surrounds the whole signal path and the feedback resistor and capacitor would bridge over the guard ring.

Thank you

Andreas

Hi Andreas,

I have the same doubt you did some time ago. How did you solve your problem?

Thank you

Best regards

Mariana

Andreas and Mariana,

The figures in the OPA129 data sheet assume that the photodiode is connected with an air-wired connection to the input summing junction (inverting input) pad inside the guard ring. The feedback resistor would also connect inside the guard ring.

If you have longer connections to be made, the guard ring can be extended around summing junction trace. Generally, however, you want to minimize the size of the summing junction trace to minimize leakage current and stray capacitance.

Regards, Bruce.

I extended the guard ring a bit to include the solder pad for the photodiode and used a larger package for my feedback resistor and capacitor to bridge over the ring.

In the past two days I've been reading every document possible about Guarding and i always get to the same point with the same questions (until i got here and realized Andreas and Mariana had the exact same problem).

If im getting it correctly, Guarding should be done in both sides of the PCB for a DIP package, however, if im using SMT, the guard should be done in just one layer, and there's when i get the problem, because i cant associate figures 2, 3, and 4 of the datasheet if one is trying to work with surface mount resistors and capacitors.

Please correct me if im wrong Bruce, but if i want to use the guard ring as suggested, do i have to use axial elements?  

Thank you very  much in advance and best regards.

JACR,

The femtoampere designs are tricky.  You need to think about where leakage can occur and try your best to block it. It sounds as if you are thinking of the right issues.

Yes, you should guard the bottom side of the circuit board if you are using through-hole components. The bottom side has equal risk of leakage paths. With surface mount components, the bottom side is not so important. The board itself is a very good insulator.

I assume your point on the surface-mount components is that it is difficult or impossible to run a guard trace between the ends of these devices. I have not searched but there may be special form factors for high value resistors. You might want to do some searching or calling vendors. I would guess that capacitors are going to be more difficult.

Very low leakage designs often use through-hole components, soldering them down on surface-mount pads. Flying leads and "air wiring" are common. Some companies used DIP package op amps and bent the critical input pin up for air wiring. You may need to get creative, depending on the level of performance you are trying to achieve. Some companies go to no special efforts including no guarding. It may depend on the environment your circuit will live and its expected lifetime. A well-cleaned circuit board in a clean, dry, well-sealed enclosure may need little special attention. At the other extreme, have you ever looked inside an old CRT TV or monitor (I guess they're all old now). Yikes! The high voltage supply and ventilation are a dust magnet.

Regards, Bruce.

If I'm getting it correctly, guarding should be done in both sides of the PCB for a DIP package, however, if I'm using SMT, the guard should be done in just one layer, and there's when I get the problem, because I can't associate figures 2, 3, and 4 of the datasheet if one is trying to work with surface mount resistors and capacitors.

On my schematics, the ring is drawn right through the middle of the components to emphasize that. I generally run a 10 mil/0.25 mm trace under an 0805 component. The fun comes when the other lead of the component is also high-impedance; then you need to guard each pin separately! I can (by pushing the pads a bit further apart than I usually do) make a 40 mil/1 mm gap under an 0805, and use 8 mil/0.20 mm trace and space.

Ty very much Bruce and Cypher but now i have the last question

If im not making a sub-picoampere measurement, it is necessary to make guard rings?  (im not taking the easy way out here, i only think it would be possible that i will increase the problem somehow)

I need to make two precision measurements, one with the OPA129 (i will use it as an electrometer to measure the voltage on a reference electrode and i will use the guard ring there) and another with the OPA140 (I-V converter with Rf=10Mohm), but the full scale would be 1uA on a 16bits system (LSB of 15.26pA), so im asking if the guard would be necessary on this latter case.

Thank you very much and once again best regards

JACR,

I assume that by "reference electrode" you are referring to some type of electro-chemical cell. These devices behave as a very weak battery. Some of these cells can be charged by the small current from input bias current of the op amp and/or leakage current. This may cause inaccuracy or damage to the cell. It all depends on the characteristics of the cell as to how sensitive it may be. This is the reason that very low input bias current op amps are often chosen for this application.

In your second application, one LSB of 15pA does not seem to be a super sensitive situation. If your circuit board is clean and will remain clean over the life of the product, it probably would not require a guard ring.

Regards, Bruce.

Tera is the reciprocal of pico, so if your signal trace is adjacent to a supply trace 10 V away, even a clean 1 TΩ board will give you 10 pA of leakage current, which is dangerously close to your LSB. My rule of thumb would want all adjacent traces within 100 mV to achieve that 10 pA limit, which requires a guard ring, although not a very careful one.

Here's an Analog devices paper on PCB layout, including guard rings.

In summary, you can probably barely get away without guard rings, but if it were me, I'd add them.

Edit: That first linked paper (Microchip AN1258) is interesting, but its guard ring illustrations are all wrong. Do not use the layouts from its figures 25 through 28, except as something to point at and laugh. The guard ring needs to be driven by a low-impedance source that is not the signal input. Which means generally the op-amp input that is not the circuit input. They have it exactly backwards!

Thank you very much, your answers made it all clear.

Yes Bruce, the reference electrode refers to a part of the electrochemical cell and im using the OPA129 as an electrometer because i don't want to polarize the electrode with the Ibias.

Very interesting article about SIR (i have never read about that before to be honest) but i already read the ADI article in my search for the "guard ring" :)

Your appreciation about the LSB is quite precise, because im working with +/- 15V supplies, so, i will probably have  15 pA of leakage current between adjacent pins 

Best regards and thanks a lot for your time and cooperation